1. Field of the Invention
The present invention is directed generally to an exposure method and an exposure apparatus for use to form a pattern of a semiconductor integrated circuit, or a liquid crystal device, or the like, and more particularly, to a projection exposure method and a projection exposure apparatus which are employed in a lithography process for liquid crystal elements and semiconductor memory cells having regular hyperfine patterns.
2. Related Background Art
A method of transferring mask patterns on a substrate typically by the photolithography method is adopted in manufacturing semiconductor memories and liquid crystal elements. In this case, the illumination light such as ultra-violet rays for exposure strikes on the substrate having its surface formed with a photosensitive resist layer through a mask formed with the mask patterns. The mask patterns are thereby photo-transferred on the substrate. In a projection exposure apparatus (for example, a stepper), the image of a circuit pattern drawn on the mask so as to be transferred is projected on the surface of the substrate (wafer) via a projection optical system so as to be imaged.
The typical hyperfine mask patterns of the semiconductor memory and the liquid crystal element can be conceived as regular grating patterns arrayed vertically or horizontally at equal spacings. Formed, in other words, in the densest pattern region in this type of mask patterns are the grating patterns in which equally-spaced transparent lines and opaque lines, formable on the substrate, for attaining the minimum line width are arrayed alternately in X and/or Y directions. On the other hand, the patterns having a relatively moderate degree of fineness are formed in other regions. In any case, the oblique patterns are exceptional.
A typical material for the photosensitive resist exhibits a non-linear photosensitive property. A chemical variation thereof quickly advances on giving an acceptance quantity greater than a certain level. If smaller than this level, however, no chemical variation advances. Hence, there exists a background wherein if a difference in light quantity between a light portion and a shade portion is sufficiently secured with respect to a mask pattern projected image on the substrate, a desired resist image according to the mask patterns can be obtained even when a boundary contrast between the light portion and the shade portion is somewhat low.
In recent years, a projection exposure apparatus such as a stepper, etc. for transferring the mask pattern on the substrate by reductive projection has been often employed with a hyperfiner pattern construction of the semiconductor memory and the liquid crystal element. Special ultra-violet rays having a shorter wavelength and narrower wavelength distributing width are employed as illumination light for exposure. The reason why the wavelength distribution width is herein narrowed lies in a purpose for eliminating a deterioration in quantity of the projected image due to a chromatic aberration of the projection optical system of the projection exposure apparatus. The reason why the shorter wavelength is selected lies in a purpose for improving the contrast of the projected image. Shortening of the wavelength of the illumination light induces a limit in terms of constraints of lens materials and resist materials in addition to the fact that no appropriate light source exists for the much hyperfiner mask patterns required, e.g., for the projection exposure of line widths on the submicron order. This is the real situation.
In the hyperfine mask patterns, a required value of the pattern resolution line width is approximate to the wavelength of the illumination light. Hence, it is impossible to ignore influences of diffracted light generated when the illumination light penetrates the mask patterns. It is also difficult to secure a sufficient light-and-shade contrast of the mask pattern projected image on the substrate. In particular, the light-and-shade contrast at the pattern line edges remarkably declines.
More specifically, respective diffracted light components, a 0th-order diffracted light component, (±) primary diffracted light components and those greater than (±) secondary diffracted light components that are generated at respective points on the mask patterns due to the illumination light incident on the mask from above pass through the projection optical system. These light components are converged again at the respective points on the substrate conjugate these points, thereby forming the image. However, the (±) primary diffracted light components and those larger than the (±) secondary diffracted light components have a much larger diffraction angle than that of the 0th-order diffracted light component with respect to the hyperfiner mask patterns and are therefore incident on the substrate at a shallower angle. As a result, a focal depth of the projected image outstandingly decreases. This causes a problem in that a sufficient exposure energy can not be supplied only to some portions corresponding to a part of thickness of the resist layer.
It is therefore required to selectively use the exposure light source having a shorter wavelength or the projection optical system having a larger numerical aperture in order to transfer the hyperfiner patterns. As a matter of course, an attempt for optimizing both of the wavelength and the numerical aperture can be also considered. Proposed in Japanese Patent Publication. No. 62-50811 was a so-called phase shift reticle in which a phase of the transmitted light from a specific portion among the transmissive portions of reticle circuit patterns deviates by π from a phase of the transmitted light from other transmissive portions. When using this phase shift reticle, the patterns which are hyperfiner than in the prior art are transferable.
In the conventional exposure apparatus, however, it is presently difficult to provide the illumination light source with a shorter wavelength (e.g., 200 nm or under) than the present one for the reason that there exists no appropriate optical material usable for the transmission optical member.
The numerical aperture of the projection optical system is already approximate to the theoretical limit at the present time, and a much larger numerical aperture can not be probably expected.
Even if the much larger numerical aperture than at present is attainable, a focal depth expressed by ±λ/2NA2 is abruptly reduced with an increase of the numerical aperture. There becomes conspicuous the problem that the focal depth needed for an actual use becomes smaller and smaller. On the other hand, a good number of problems inherent in the phase shift reticle, wherein the costs increase with more complicated manufacturing steps thereof, and the inspecting and modifying methods are not yet established.
In an irradiation optical system for irradiating the reticle with light, an optical integrator such as a fly-eye type optical integrator (a fly-eye lens) and a fiber is used so as to uniform the distribution of the intensities of irradiation light with which the surface of the reticle is irradiated. In order to make the aforesaid intensity distribution uniform optimally, a structure which employs the fly-eye lens is arranged in such a manner that the reticle-side focal surface (the emission side) and the surface of the reticle (the surface on which the pattern is formed) hold a substantially Fourier transform relationship. Also the focal surface adjacent to the reticle and the focal surface adjacent to the light source (the incidental side) hold the Fourier transform relationship. Therefore, the surface of the reticle, on which the pattern is formed, and the focal surface of the fly-eye lens adjacent to the light source (correctly, the focal surface of each lens of the fly-eye lens adjacent to the light source) hold an image formative relationship (conjugated relationship). As a result of this, irradiation light beams from respective optical elements (a secondary light source image) of the fly-eye lens are added (superposed) because they pass through a condenser lens or the like so that they are averaged on the reticle. Hence, the illuminance uniformity on the reticle can be improved. Incidentally, there has been disclosed an arrangement capable of improving the illuminance uniformity in U.S. Pat. No. 4,497,015 in which two pairs of optical integrators are disposed in series.
In a conventional projection exposure apparatus, the light quantity distribution of irradiation beams to be incident on the optical integrator, such as the aforesaid fly-eye lens, has been made to be substantially uniform in a substantially circle area (or in a rectangular area), the center of which is the optical system of the irradiation optical system.
FIG. 54 illustrates a schematic structure of a conventional projection exposure apparatus (stepper) of the above described type. Referring to FIG. 54, irradiation beams L340 pass through a fly-eye lens 241c, a spatial filter (an aperture diaphragm) 205a and a condenser lens 208 so that a pattern 210 of a reticle 209 is irradiated with the irradiation beams L340. The spatial filter 205a is disposed on, or adjacent to a Fourier transform surface 217 (hereinafter abbreviated to a “pupil surface or plane”) and also referred to as a Fourier transform plane with respect to the reticle side focal surface 614c of the fly-eye lens 241c, that is, with respect to the reticle pattern 210. Furthermore, the spatial filter 205a has a substantially circular opening centered at a point on optical axis AX of a projection optical system 211 so as to limit a secondary light source (plane light source) image to a circular shape. The irradiation light beams, which have passed through the pattern 210 of the reticle 209, are imaged on a resist layer of a wafer 213 via the projection optical system 211. In the aforesaid structure, the numerical aperture of the irradiation optical system (241c, 205a and 208) and the reticle-side numerical aperture formed in the projection optical system 211, that is σ value is determined by the aperture diaphragm (for example, by the diameter of an aperture formed in the spatial filter 205a), the value being 0.3 to 0.6 in general.
The irradiation light beams L340 are diffracted by the pattern 210 patterned by the reticle 209 so that 0-order diffracted light beam Do, +1-order diffracted light beam Dp and −1-order diffracted light beam Dm are generated from the pattern 210. The diffracted light beams Do, Dp and Dm, thus generated, are condensed by the projection optical system 211 so that interference fringes are generated. The interference fringes, thus generated, correspond to the image of the pattern 210. At this time, angle θ (reticle side) made by the 0-order diffracted light beam Do and ±1-order diffracted light beams Dp and Dm is determined by an equation expressed by sin θ=λ/P (λ: exposure wavelength and P: pattern pitch).
It should be noted that sin θ is enlarged in inverse proportion to the length of the pattern pitch, and therefore if sin θ has become larger than the numerical aperture (NAR) formed in the projection optical system 211 adjacent to the reticle 209, the ±1-order diffracted light beams Dp and Dm is limited by the effective diameter of a pupil (a Fourier transform surface) 212 in the projection optical system 211. As a result, the ±1-order diffracted light beams Dp and Dm cannot pass through the projection optical system 211. At this time, only the 0-order diffracted light beam Do reaches the surface of the wafer 213 and therefore no interference fringe is generated. That is, the image of the pattern 210 cannot be obtained in a case where sin θ>NAR. Hence, the pattern 210 cannot be transferred to the surface of the wafer 213.
It leads to a fact that pitch P, which holds the relationship sin Θ=λ/P≅NAR, has been given by the following equation.P≅λ/NAR
Therefore, the minimum pattern size becomes about 0.5 λ/NAR because the minimum pattern size is the half of the pitch P. However, in the actual photolithography process, some considerable amount of focal depth is required due to an influence of warp of the wafer, an influence of stepped portions of the wafer generated during the process and the thickness of the photoresist. Hence, a practical minimum resolution pattern size is expressed by k·λ/NAR, where k is a process factor which is about 0.6 to 0.8. Since the ratio of the reticle side numerical aperture NAR and the wafer side numerical aperture NAW is the same as the imaging magnification of the projection optical system, the minimum resolution size on the reticle is k·λ/NAR and the minimum pattern size on the wafer is k·λ/NAW=k·λ/B·NAR (where B is an imaging magnification (contraction ratio)).
Therefore, a selection must be made whether an exposure light source having a shorter wavelength is used or a projection optical system having a larger numerical aperture is used in order to transfer a more precise pattern. It might, of course, be considered feasible to study to optimize both the exposure wavelength and the numerical aperture.
However, as pointed out earlier, it is so far difficult for the projection exposure apparatus of the above described type to shorten the wavelength of the irradiation light source (for example, 200 nm or shorter) because a proper optical material to make a transmissive optical member is not present and so forth. Furthermore, the numerical aperture formed in the projection optical system has approached its theoretical limit at present and therefore it is difficult to further enlarge the numerical aperture. Even if the numerical aperture can be further enlarged, the focal depth expressed by ±λ/2NA2 rapidly decreases with an increase in the numerical aperture, causing a critical problem to take place in that the focal depth required in a practical use further decreases.
As pointed out earlier, by using a phase shift reticle of the type described above, a further precise pattern can be transferred. However, the phase shift reticle has a multiplicity of unsolved problems because of a fact that the cost cannot be reduced due to its complicated manufacturing process and inspection and modification methods have not been established even now.
Hence, an attempt has been made as projection exposure technology which does not use the phase shift reticle and with which the transference resolving power can be improved by modifying the method of irradiating the reticle with light beams. One irradiation method of the aforesaid type is a so-called annular zone irradiation method, for example; arranged in such a manner that the irradiation light beams which reach the reticle 209 are given a predetermined inclination by making the spatial filter 205a shown in FIG. 54 an annular opening so that the irradiation light beams distributed around the optical axis of the irradiation optical system are cut on the Fourier transform surface 217.
In order to establish projection exposure having a further improved resolving power and a larger focal depth, an inclination irradiation method or a deformed light source method has been previously disclosed in PCT/JP91/01103 (filed on Aug. 19, 1991). The aforesaid irradiation method is arranged in such a manner that a diaphragm (a spatial filter) having a plurality (two or four) openings, which are made to be eccentric with respect to the optical axis of the irradiation optical system by a quantity corresponding to the precision (the pitch or the like) of the reticle pattern, is disposed adjacent to the emission side focal surface of the fly-eye lens so that the reticle pattern is irradiated with the irradiation light beams from a specific direction while inclining the light beams by a predetermined angle.
However, the above mentioned inclination irradiation method and the deformed light source method have a problem in that it is difficult to realize a uniform illuminance distribution over the entire surface of the reticle because the number of effective lens elements (that is, the number of secondary light sources capable of passing through the spatial filter) decreases and therefore an effect of making the illuminance uniform on the reticle deteriorates. What is worse, the light quantity loss is excessively large in the system which has a member, such as the spatial filter, for partially cutting the irradiation light beams. Therefore, the illumination intensity (the illuminance) on the reticle or the wafer can, of course, deteriorate excessively, causing a problem to take place in that the time taken to complete the exposure process becomes long with the deterioration in the irradiation efficiency. Furthermore, a fact that light beams emitted from the light source concentrically pass through the Fourier transform plane in the irradiation optical system will cause the temperature of a light shielding member, such as the spatial filter, to rise excessively due to its light absorption and a measure (air cooling or the like) must be taken to prevent the performance deterioration due to change in the irradiation optical system caused from heat.
In a case where a diaphragm of the aforesaid type is disposed adjacent to the emission side focal surface of the fly-eye lens, some of the secondary light source images formed by a plurality of the lens elements are able to superpose on the boundary portion between the light transmissive portion of the diaphragm and the light shielding portion of the same. This means a fact that the secondary light source image adjacent to the aforesaid boundary portion is shielded by the diaphragm or the same passes through the boundary portion on the contrary. That is, an unstable factor, such as the irradiation light quantity, is generated and another problem arises in that the light quantities of the light beams emitted from the aforesaid diaphragm and that are incident on the reticle become different from one another. Furthermore, in the inclination irradiation method, the positions of the four openings (in other words, the light quantity distribution in the Fourier transform plane) must be changed in accordance with the degree of precision of the reticle pattern (the line width, or the pitch or the like). Therefore, a plurality of diaphragms must be made to be exchangeable in the irradiation optical system, causing a problem to arise in that the size of the apparatus is enlarged.
When a secondary light source formed on the reticle side focal surface of the fly-eye lens is considered in a case where the light source comprises a laser such as an excimer laser having a spatial coherence, the irradiation light beams corresponding to the lens elements have some considerable amount of coherence from each other. As a result, random interference fringes (speckle interference fringes) are formed on the surface of the reticle or the surface of the wafer which is in conjugate with the surface of the reticle, causing the illuminance uniformity to deteriorate. When its spatial frequency is considered here, a Fourier component corresponding to the minimum interval between the lens elements is present in main. That is, the number of combinations of light beams contributing to the interference is the largest. Therefore, fringes having a relatively low frequency (having a long pitch) in comparison to the limit resolution and formed to correspond to the configuration direction of the lens elements are observed on the surface of the reticle or the surface of the wafer. Although the formed interference fringes have low contrast because the KrF excimer laser has a relatively low spatial coherence, the interference fringe acts as parasite noise for the original pattern. The generation of the interference fringes causes a problem when the illuminance uniformity, which will be further required in the future, is improved. In the case where the annular zone irradiation method is considered, the aforesaid noise concentrically superposes in the vicinity of the limit resolution, and therefore the influence of the noise is relatively critical in comparison to the ordinary irradiation method (see FIG. 54).
Disclosed, on the other hand, in U.S. Pat. No. 4,947,413 granted to T. E. Jewell et al is the projection lithography method by which a high contrast pattern projected image is formed with a high resolving power on the substrate by making the 0th-order diffracted light component coming from the mask patterns and only one of the (+) and (−) primary diffracted light components possible of interference by utilizing a spatial filter processing within the Fourier transform plane in the projection optical system by use of an off-axis illumination light source. Based on this method, however, the illumination light source has to be off-axis-disposed obliquely to the mask. Besides, the 0th-order diffracted light component is merely interfered with only one of the (+) and (−) primary diffracted light components. Therefore, the light-and-shade contrast of edges of the pattern image is not yet sufficient, the image being obtained by the interference due to unbalance in terms of a light quantity difference between the 0th-order diffracted light component and the primary diffracted light component.